AI LLM

The AI LLM Step allows you to leverage the power of Generative AI as part of your Processes. To use it, you must first configure an Integration with a Generative AI service. After an appropriate Integration has been set up and configured, you'd need to configure two prompts: the system prompt and the user prompt. The system prompt is used to provide the language model with general instructions and directives. For example, if you're asking the model to behave as a helpful automated assistant, you could specify how you'd like it to carry out its job and what tone of voice to use. The user prompt is more about the task at hand. Building on the previous example, you could ask the LLM to read a long chat history and summarize the key points. Or you could ask it to interpret natural language in order to determine how to proceed with the Process. The possibilities are infinite!

    "type": "llm",
    "name": "",
    "publishProperties": {
      "allowedTokens": 500,
      "temperature": 0.7,
      "responseType": "json_object"
    },
    "messageBody": {
      "systemPrompt": "You are a highly intelligent AI assistant that generates creative and structured responses in JSON format. Always ensure the response is accurate, informative, and easy to understand.",
      "userPrompt": "Generate a random JSON object that describes a futuristic city, including details like the city's name, population size, main technology used, most popular form of transportation, and one unique cultural trait."
    }

Input

Field	Description	Mandatory
type	Always llm.	✔
name	The name of the integration.	✔
publishPropeties.allowedTokens	The maximum number of tokens that can be generated.
publishPropeties.temperature	The sampling temperature for text generation. Can be any number between 0-2, defaults to 1. Higher means more random.
publishPropeties.responseType	If json_object is specified, converts the generated JSON into an object.
messageBody.systemPrompt	The system prompt to use for this Step.
messageBody.userPrompt	The user prompt to use for this Step.

tip

You can weave in data from the rest of the Process by using the {{ }} syntax.

Output

Field	Description
result	The generated text or JSON.